Image Editing & Inpainting¶
Image & 2D Art
Open Source¶
-
HiFi-Inpaint: Towards High-Fidelity Reference-Based Inpainting for Generating Detail-Preserving Human-Product Images Discovered by AI
2026
A high-fidelity reference-based inpainting framework designed to generate detail-preserving human-product images. It introduces Shared Enhancement Attention (SEA) to refine fine-grained product features and a Detail-Aware Loss (DAL) to enforce precise pixel-level supervision.
detail-preservinge-commerceinpaintingreference-basedGitHub · arXiv · Hugging Face · Website
-
Moebius Discovered by AI
2026
Moebius is a highly efficient, 0.2B lightweight image inpainting framework that achieves performance comparable to 10B-level foundation models. It reconstructs the diffusion backbone by introducing a dual-track sparse transformer and a dynamic feature distillation mechanism to significantly reduce computational costs.
diffusioninpaintinglightweight -
ACE++ Discovered by AI
2025
An instruction-based diffusion framework developed by Alibaba Tongyi Lab that unifies reference image generation, local editing, and controllable generation. It improves upon the Long-context Condition Unit (LCU) and utilizes a two-stage training scheme to enable high-quality context-aware content filling and precise instruction following.
diffusionimage-editinginpaintinginstruction-following -
2025
A 12B flow-matching model for instruction-based image editing that preserves characters and objects across iterative edits, supporting local and global edits plus style and character reference.
character-consistencyfluxinstruction-editingGitHub · arXiv · Hugging Face
-
ICEdit: Enabling Instructional Image Editing with In-Context Generation in Diffusion Transformers Discovered by AI
2025
ICEdit is an in-context image editing framework built on large-scale Diffusion Transformers (DiTs). It enables precise, instruction-based image editing by leveraging the model's inherent generation and understanding capabilities without requiring architectural modifications or extensive fine-tuning.
diffusion-transformerimage-editingin-context-learning -
InfiniteYou Discovered by AI
2025
InfiniteYou is an identity-preserving image editing and photo recrafting framework built on advanced Diffusion Transformers (DiTs) like FLUX. It introduces InfuseNet to inject identity features into the DiT backbone, enabling flexible, high-fidelity, and aesthetically pleasing character editing while maintaining strict identity similarity.
diffusion-transformeridentity-preservingimage-editing -
Inpaint4Drag Discovered by AI
2025
Inpaint4Drag is a drag-based image editing framework that decomposes the drag process into pixel-space bidirectional warping and generative image inpainting. It repurposes off-the-shelf inpainting models to achieve precise, geometry-consistent, and interactive object relocation and deformation.
drag-editingimage-manipulationinpainting -
MagicQuill V2 Discovered by AI
2025
MagicQuill V2 introduces a layered composition paradigm to generative image editing, deconstructing creative intent into distinct, controllable visual cues (Content, Spatial, Structural, and Color). This allows users to perform precise, interactive image editing with granular control over object placement, shape, and appearance.
image-editinginteractive-editinglayered-composition -
2025
A unified model that handles text-to-image generation and instruction-based editing in one architecture, accepting interleaved text-and-image prompts for compositional edits.
instruction-editingmultimodalunified -
PixelHacker Discovered by AI
2025
A diffusion-based image inpainting model that introduces Latent Categories Guidance (LCG) to maintain structural and semantic consistency. It achieves state-of-the-art performance on benchmarks like Places2, CelebA-HQ, and FFHQ.
diffusioneditinginpainting -
Step1X-Edit Discovered by AI
2025
Step1X-Edit is a state-of-the-art image editing framework that leverages Multimodal Large Language Models (MLLMs) to parse instructions and generate editing tokens, which are then decoded into images using a Diffusion Transformer (DiT) backbone. It achieves performance comparable to closed-source models like GPT-4o and Gemini 2 Flash.
diffusion-transformerimage-editingmllm -
Token Painter Discovered by AI
2025
A training-free text-guided image inpainting method based on Mask Autoregressive Models (MAR) that selectively generates masked tokens while preserving the unmasked background.
autoregressiveimage-editinginpaintingtraining-free -
TurboFill Discovered by AI
2025
A fast image inpainting model that adapts a few-step text-to-image diffusion model (like DMD2) with an inpainting adapter, achieving high-quality, structurally consistent, and visually harmonious inpainting in just 4 steps.
cvpr2025diffusionfew-stepinpainting -
AnyEdit: Mastering Unified High-Quality Image Editing for Any Idea Discovered by AI
2024
A comprehensive multi-modal instruction-based image editing framework trained on a high-quality dataset of 2.5 million editing pairs across 20+ editing types.
diffusionimage-editinginstruction-guided -
AnyText2 Discovered by AI
2024
A multilingual visual text generation and editing model that enables precise control over text attributes (such as font, color, size, and orientation) while seamlessly inpainting or editing text within natural scene images.
diffusioninpaintingmultilingualtext-editing -
BrushEdit Discovered by AI
2024
BrushEdit is an all-in-one image inpainting and editing framework that addresses the limitations of inversion-based and instruction-based methods. It allows users to perform precise local editing, object addition, and removal with high fidelity and structural consistency.
diffusionimage-editinginpainting -
BrushNet Discovered by AI
2024
A plug-and-play image inpainting model that introduces a decomposed dual-branch diffusion architecture, separating masked image features and noisy latents to preserve image quality and semantic consistency.
diffusioninpaintingplug-and-play -
CAT-Diffusion Discovered by AI
2024
A cascaded transformer-diffusion framework for text-guided object inpainting that pre-inpaints target objects in an auxiliary multi-modal feature space (CLIP) before generating high-fidelity visual details.
clipdiffusionimage-editinginpainting -
ComfyUI Inpaint Nodes Discovered by AI
2024
A collection of custom nodes for ComfyUI designed to improve inpainting and outpainting workflows, supporting models like Fooocus inpaint for SDXL, LaMa, and MAT.
comfyuiinpaintinglamaoutpainting -
DesignEdit Discovered by AI
2024
A training-free multi-layered latent diffusion framework that unifies various spatial-aware image editing tasks (such as inpainting, outpainting, moving, and resizing) by decomposing the image into layers.
diffusionimage-editinginpaintingoutpainting -
DiffEditor Discovered by AI
2024
A diffusion-based image editing framework that improves fine-grained editing accuracy and flexibility. It integrates image prompts, a combined SDE-ODE formulation, regional score-based gradient guidance, and a time travel strategy to enable precise local modifications.
diffusiondragediting -
Diffree Discovered by AI
2024
A text-guided, shape-free object inpainting model that seamlessly inserts new objects into images without requiring user-defined masks. It automatically predicts the appropriate spatial location and mask for the new object while maintaining background consistency.
diffusioninpaintingobject-addition -
Edicho Discovered by AI
2024
A training-free image editing framework based on diffusion models that enables consistent editing across in-the-wild images by utilizing explicit image correspondence to direct the editing process.
diffusioniccv2025image-editingtraining-free -
FlowEdit Discovered by AI
2024
A training-free and inversion-free text-based image editing method that utilizes pre-trained text-to-image flow models like Stable Diffusion 3 and FLUX, constructing a direct ordinary differential equation (ODE) to map source and target distributions.
flow-matchingimage-editinginversion-free -
FLUX ControlNet Inpainting Discovered by AI
2024
An advanced inpainting ControlNet checkpoint designed for the FLUX.1-dev model, enabling high-quality, context-aware image completion and local editing.
controlnetfluxinpainting -
MagicQuill Discovered by AI
2024
An intelligent interactive image editing system that integrates a multimodal large language model (MLLM) to interpret user brush strokes and text prompts, enabling precise and flexible editing operations such as object insertion, removal, and color alteration with minimal user input.
image-editinginteractivemllm -
Paint by Inpaint: Learning to Add Image Objects by Removing Them First Discovered by AI
2024
A mask-free, instruction-following image editing framework that trains a diffusion model to add objects to images by learning to inverse the inpainting process.
diffusionimage-editinginpainting -
StrDiffusion: Structure Matters: Tackling the Semantic Discrepancy in Diffusion-Based Image Inpainting Discovered by AI
2024
A structure-guided diffusion model for image inpainting that reformulates the conventional texture denoising process under the guidance of an auxiliary sparse structure to resolve semantic discrepancies.
diffusionimage-editinginpainting -
UltraEdit: Instruction-based Fine-Grained Image Editing at Scale Discovered by AI
2024
A large-scale dataset and framework for instruction-based fine-grained image editing, supporting both free-form and region-based editing with high-quality automatically produced region annotations.
datasetdiffusionimage-editinginstruction-based -
AnyDoor: Zero-shot Object-level Image Customization Discovered by AI
2023
A diffusion-based image generator designed for zero-shot object-level image customization, allowing users to seamlessly teleport target objects into new scenes at specified locations with desired shapes.
diffusionimage-editinginpainting -
AnyText Discovered by AI
2023
AnyText is a specialized visual text generation and editing tool designed to render accurate and coherent multilingual text within images. It supports text inpainting and editing, allowing users to seamlessly replace, modify, or add text in specific regions of an image.
inpaintingmultilingualtext-editing -
DragDiffusion Discovered by AI
2023
An interactive point-based image editing framework that extends drag-style pixel manipulation to diffusion models, achieving precise control over object poses, shapes, and layouts.
drag-basedimage-editinginteractive -
HD-Painter Discovered by AI
2023
HD-Painter is a training-free approach for high-resolution and prompt-faithful text-guided image inpainting. It introduces a prompt-aware attention guidance mechanism and a specialized scale-aware sampling strategy to align inpainted regions with complex user prompts.
diffusionhigh-resolutioninpainting -
Inpaint Anything Discovered by AI
2023
A mask-free image inpainting pipeline that combines the Segment Anything Model (SAM) with diffusion-based inpainting models to enable click-and-fill object removal, replacement, and background editing.
inpaintingsamsegment-anything -
Inpaint-Web Discovered by AI
2023
A free and open-source inpainting and image-upscaling tool powered by WebGPU and WebAssembly (WASM) that runs entirely client-side in the browser.
browser-toolinpaintingwasmwebgpu -
LEDITS++ Discovered by AI
2023
An efficient and precise text-guided image editing method that supports multiple simultaneous edits, requiring no fine-tuning while maintaining high fidelity to the original image.
diffusionimage-editingtext-guided -
MagicBrush Discovered by AI
2023
A large-scale, manually annotated dataset and benchmark for instruction-guided real image editing, covering single-turn, multi-turn, mask-provided, and mask-free editing scenarios.
datasetimage-editinginstruction-guided -
MagicRemover Discovered by AI
2023
A tuning-free text-guided image inpainting method that utilizes attention guidance within the diffusion model's sampling process to precisely erase objects and restore occluded content.
diffusioninpaintingtuning-free -
MasaCtrl Discovered by AI
2023
A tuning-free mutual self-attention control method that enables consistent image synthesis and non-rigid image editing by querying self-attention maps from source images.
image-editingself-attentiontuning-free -
Pix2Pix-Zero Discovered by AI
2023
A zero-shot image-to-image translation method that preserves the structure of the original image by finding edit directions in the text embedding space without requiring manual prompts or training.
image-editingimage-to-imagezero-shot -
PowerPaint Discovered by AI
2023
A versatile image inpainting model that supports text-guided object inpainting, object removal, image outpainting, and shape-guided object inpainting within a single unified framework.
inpaintingobject-removaloutpainting -
SketchRefiner Discovered by AI
2023
An interactive image inpainting framework that utilizes user-drawn sketches to guide the image restoration process. It introduces a sketch refinement stage to bridge the gap between rough, hand-drawn sketches and the precise structural inputs required by inpainting models.
inpaintinginteractivesketch-guided -
SmartEdit: Exploring Complex Instruction-based Image Editing with Multimodal Large Language Models Discovered by AI
2023
An instruction-based image editing framework that integrates Multimodal Large Language Models (MLLMs) to enhance the understanding and reasoning capabilities for complex editing instructions.
diffusionimage-editinginstruction-basedmllm -
DiffEdit Discovered by AI
2022
A semantic image editing method that automatically generates mask guidance from text queries using diffusion models, enabling precise local edits while preserving the rest of the image.
diffusionimage-editingmask-guidance -
InstructPix2Pix Discovered by AI
2022
An instruction-guided image editing model trained on paired datasets generated by GPT-3 and Stable Diffusion, allowing users to edit images using natural language instructions.
diffusionimage-editinginstruction-guided -
IOPaint Discovered by AI
2022
A popular open-source web-based image inpainting and editing tool powered by state-of-the-art AI models, supporting object removal, erasing, and stable diffusion-based replacement.
inpaintingobject-removalweb-ui -
Paint-by-Example Discovered by AI
2022
An exemplar-guided image editing framework that leverages self-supervised training and an information bottleneck to edit images using an exemplar image instead of text prompts, enabling precise control over object insertion and style.
diffusionexemplar-guidedimage-editing -
Plug-and-Play Diffusion Features Discovered by AI
2022
A framework that extracts and injects spatial features and self-attention maps from a guidance image into a text-to-image diffusion process, enabling high-fidelity image-to-image translation.
diffusionimage-editingimage-to-image -
Prompt-to-Prompt Discovered by AI
2022
A training-free image editing framework that manipulates cross-attention maps in text-to-image diffusion models to achieve precise semantic edits while preserving the original layout.
cross-attentionimage-editingtraining-free -
RePaint Discovered by AI
2022
A classic denoising diffusion probabilistic model (DDPM) based inpainting approach that modifies the reverse diffusion process to handle extreme and free-form masks. It leverages a pretrained unconditional DDPM as a generative prior, requiring no task-specific training.
ddpmdiffusioninpainting -
LaMa Discovered by AI
2021
A highly robust image inpainting method based on fast Fourier convolutions (FFCs) that excels at filling large missing areas and complex geometric structures in high-resolution images.
fourier-convolutionhigh-resolutioninpainting -
SDEdit Discovered by AI
2021
A foundational image synthesis and editing method that adds noise to a guide image and denoises it using stochastic differential equations to balance realism and faithfulness to user inputs.
diffusionimage-editingsde
Proprietary / Industry Tools¶
-
Adobe Firefly / Photoshop Generative Fill Discovered by AI
2023
Adobe's flagship generative AI tool integrated directly into Photoshop, allowing artists to add, extend, or remove content from images using simple text prompts. It is widely used in professional game and anime production for concept art iteration, background cleanup, and asset generation.
commercialinpaintingoutpaintingphotoshop -
Adobe Firefly Generative Fill Discovered by AI
2023
Adobe's commercial generative AI tool integrated directly into Photoshop, allowing artists to add, extend, or remove content from images using simple text prompts with professional-grade control and licensing safety.
commercialinpaintingoutpaintingphotoshop -
Leonardo AI Canvas Discovered by AI
2023
Leonardo AI is a popular generative art platform featuring the AI Canvas, an interactive workspace designed for precise image editing, inpainting, outpainting, and style-consistent asset generation. It is widely used by game and anime studios for rapid prototyping and asset refinement.
image-editinginpaintingoutpainting -
Clipdrop Discovered by AI
2020
An AI-powered visual editing suite by Stability AI/Jasper that offers advanced tools for object removal (Cleanup), image outpainting (Uncrop), relighting, and background replacement.
cleanupimage-editinginpaintingoutpainting -
Photoroom Discovered by AI
2020
Photoroom is a proprietary AI-powered photo editing platform specializing in background removal, object removal, and high-quality image inpainting. It provides web, mobile, and batch API solutions tailored for rapid asset creation, e-commerce, and marketing production.
background-removalinpaintingobject-removalproprietary
:quality(80):no_upscale())